nielsmadan
- 63 skills
- 0 followers
- 1 repo stars
- 4 hours ago last updated
- ▌ Crew · nielsmadan bundleRun one lead session that works with the user plus one or more implementer sessions in their own worktrees or sibling clones, connected by session messages and brief files, with every change integrated locally by fast-forward and nothing pushed. Covers setup, joining as a worker, writing and delivering briefs, the worker's implement-and-report loop, lead-side integration, and bringing every other checkout up to date without a push-and-pull round trip. Use when the user says "crew", "set up workers", "one session reviews, another implements", "hand this to dev2", "join the crew", "crew join", "integrate dev2", "bring my other checkouts up to date", "sync dev1 into dev2 without pushing"; also the hand-off layer for codebase-review --fix and release-review. Not for subagents inside one session, or for a single-session build (plan, longshot).
- ▌ QA Tour · nielsmadanPrepare a developed feature for a short interactive walkthrough, then use guide to help the user click through its main workflow themselves. Use for "qa-tour", "give me a tour of the new feature", "walk me through trying this feature", or "let me click through what we built". Defaults to staged and unstaged changes, including untracked files; accepts --staged, --unstaged, --changed, --unpushed, --all, or a feature, URL, or app. Use qa for exhaustive agent-driven testing and guide alone for general setup instructions.
- ▌ Code Tour · nielsmadanGive an interactive tour of code, starting with an overview of the changes and tracing concrete inputs through entry points, calls, state changes, and branches, pausing for questions at each step. Use for "code-tour", "give me a tour of the code", or "walk me through this code step by step". Defaults to staged changes; accepts --staged, --changed, --unstaged, --unpushed, --all, or a target. Reuses relevant QA screenshots and logs. Use qa-tour for a walkthrough the user operates in the running app, and explain for a one-shot explanation or generated documentation.
- ▌ Brainstorm · nielsmadan bundleWork out how to solve a problem the user can already name — interview them with `blind-spots` to understand it, generate structurally different approaches, validate each with an independent critic, then present the survivors with trade-offs and a recommendation. Use when the user says "brainstorm", "brainstorm approaches", "how should we solve/approach/tackle X", "what are our options", "compare approaches", "think this through with me", or wants solutions weighed before committing. Works on any topic, with codebase recon when a repo is involved; `--debate` adds external advisors for high-stakes calls. Do NOT use when stuck on what to build or what the problem is (use `ideation`), to probe an already-chosen plan (use `blind-spots`), or to pick a library or service (use `evaluate-tech`).
- ▌ Review CLI · nielsmadan bundleLanguage-agnostic review of a command-line tool's interface and behaviour — exit codes and signals, broken-pipe handling, stdout/stderr discipline, TTY detection and destructive-action gating, dry-run integrity, partial failure and idempotency, config precedence, flag and help design, and error messages that name the next action. Starts by RUNNING the binary, because most of this is invisible in the source. Applies equally to Python, Rust, Go, Node and shell-invoked tools. Auto-invoked by `code-review` on projects that ship a console entry point. Triggers "review cli", "cli review", "command line interface review", "review the CLI UX".
- ▌ Rn Upgrade · nielsmadan bundleUse when upgrading React Native or Expo SDK versions or resolving upgrade-related dependency issues. Detects Expo projects and follows SDK-compatible versions and their native-project workflow.
- ▌ Review Rust · nielsmadan bundleRust-specific code review focused on JUDGMENT-level design that clippy and the compiler cannot decide — modeling state so invalid combinations are unrepresentable, sum types that get flattened into correlated primitives, closed vocabularies carried as strings, units that live only in comments, error types that lose the cause at a boundary, ownership and trait choices, unsafe justifications, async task ownership and cancellation, and semver commitments. Deliberately does NOT duplicate clippy, and opens by checking which of its 839 lints the project actually runs. Auto-invoked by `code-review` on Rust projects. Triggers "review rust", "rust review", "clippy review", "type modeling review".
- ▌ Review Tidy · nielsmadanCoordinate comment cleanup through review-comments and fix the mechanical loose ends a change leaves behind — docs, examples and config that name a symbol, flag or path the change renamed or removed; imports, private helpers and files it orphaned; test names that no longer match; and the project's own formatter and lint autofixes. Use before a review or PR so reviewers see tidy code. code-review runs it as a preflight.
- ▌ Optimize SEO · nielsmadan bundleAudit and optimize web pages for SEO — meta tags, Open Graph, JSON-LD, structured data. Triggers "optimize SEO", "check SEO", "add meta tags", "SEO audit".
- ▌ Review React · nielsmadan bundleReact-specific code review for both react-dom and React Native, focused on JUDGMENT-level component design a linter can't decide — state modeling and placement, effects that shouldn't be effects, component boundaries and identity, render cost calibrated to whether React Compiler is on, and platform pitfalls (hydration and Server Component boundaries on the web; lists, JS-thread work and native subscriptions on React Native). Deliberately does NOT duplicate eslint-plugin-react-hooks, and first checks which of its rules actually run — the compiler-backed rules from v6 on cover much of what reviewers used to flag by hand. Auto-invoked by `code-review` on React projects. Triggers "review react", "react review", "review components", "review hooks", "review effects", "react native review".
- ▌ Review Python · nielsmadan bundlePython-specific code review focused on JUDGMENT-level design a linter and type checker can't decide — modeling state so invalid combinations are unrepresentable, keeping parsed data parsed instead of re-widening it to dicts, exception hierarchies that callers can act on, async task lifetime and cancellation, and boundary discipline for subprocess, filesystem and untrusted input. Deliberately does NOT duplicate ruff or mypy. Auto-invoked by `code-review` on Python projects. Triggers "review python", "python review", "asyncio review", "type modeling review".
- ▌ Release Review · nielsmadan bundleRun a comprehensive pre-release review of a whole product, working through every command, parameter and config surface with the user, one item at a time. Rulings are collected in an on-disk changes ledger, ruled batches are handed to parallel implementer sessions as soon as they are ready, and the programme is sequenced through code-quality review, docs review, QA and release, then harvested into the repo's docs. Resumable across sessions. Use when the user says "get this to 1.0", "release review", "final review of all commands and params", "review the whole product before release", "comprehensive review of full functionality", "pre-release review", "resume the release review", or "where are we in the review". Do NOT use for reviewing a diff or PR (use code-review), QA of one feature (use qa), a persona-based product audit (use review-product), or executing an already-written plan (use longshot).
- ▌ Review Swiftui · nielsmadan bundleReview SwiftUI presentation and interaction on macOS and other Apple platforms. Checks control semantics, navigation and presentation, adaptive layout, localization, keyboard focus, and window commands against concrete user workflows. Use for "review swiftui", "swiftui review", or reviewing a SwiftUI screen's native behavior. Auto-included by code-review for scoped SwiftUI implementation. Leaves Swift modeling, concurrency, ownership, SwiftUI state, identity, dependencies, and rendering performance to review-swift. Not a general Swift or UIKit/AppKit-only review.
- ▌ Codebase Review · nielsmadan bundleReview a whole codebase aspect by aspect, with aspects fitted to the project type (mobile, desktop, web frontend, static site, CLI, backend, library) — docs truth-check, architecture, tests, security, the type's foundations, feature deep dives, consistency, roadmap — writing a doc per aspect and a verified findings ledger that doubles as the TODO. Plans first (recon, `blind-spots`, `review-plan`, plan approval), then walks the plan with the user aspect by aspect — agents review and verify in the background, the user confirms rulings and decides judgment calls one at a time. `--auto` runs independently after the gate, with decision waves at the end. `--fix` hands settled findings through `crew` to worker sessions, integrated locally, never pushed. Use for "codebase review", "review the whole codebase", "onboard me to this repo", "deep dive this app", "audit the code base", "review and fix everything". Not for a diff or PR (code-review), a CLI release pass (release-review), or building a feature (longshot).
- ▌ Flutter Upgrade · nielsmadanFlutter upgrade workflow. Use when upgrading Flutter SDK, bumping Flutter version, migrating to a new Flutter release, or resolving breaking changes after a Flutter update.
- ▌ Review Functional · nielsmadanReview code against pragmatic functional principles — effects pushed to the edges, no hidden shared state, no mutation of a caller's data, deterministic functions, declarative transforms. Deliberately non-dogmatic: currying, point-free style, recursion and monads are out of scope and are never findings. Triggers "review functional", "functional review", "check for side effects", "is this pure", "too much mutable state", "check global state".
- ▌ Review Consistency · nielsmadanReview whether changes follow the patterns the codebase already uses — compare each added or changed construct with its existing siblings for naming, placement, mechanisms (error handling, logging, config, helpers), shape, and registration, and cite the precedent for every divergence. Triggers "review consistency", "consistency review", "does this follow existing patterns", "is this consistent with the codebase", "match the conventions", "pattern drift", "check conventions". Not for style a formatter or linter decides, or for judging whether the existing pattern is any good (use review-cleancode).
- ▌ Review Correctness · nielsmadan bundleReview code for bugs, incorrect state transitions, races, and failure handling, including closed-world branching, unknown external values, swallowed errors, masking defaults, invariants, deadlines, cancellation, and retry safety. Use for "review correctness", "review logic", "review defensive", "check exhaustiveness", "fail fast review", or "robustness review".
- ▌ QA · nielsmadan bundleExercise a developed feature through its real user or consumer interface, enumerate paths and edge/error/loading states first, and report evidence and coverage. Use for "qa", "QA the last feature", "test this feature end to end", "manual QA", or "check the user flows". Defaults to the last developed feature; accepts a feature, URL, command, app, or other scope. Covers web, native apps, CLIs, agent plugins, libraries, and backends. Use test for suite work and code-review for source review.
- ▌ Doc · nielsmadan bundleAssess documentation and run the right action: check changed code when the tree is dirty, otherwise review the whole repo for gaps, staleness, and quality. Explicit modes: --review, --update, --generate, or --session to capture durable knowledge from the current conversation or a transcript. Use for doc creation, freshness, quality, or saving occasional manual test procedures and results for future repetition; routine automated suite runs do not need test records.
- ▌ Huh · nielsmadanExpand an explanation with the surrounding context needed to understand it. Use when the user invokes `huh`, asks to explain the previous reply in more detail, or supplies text to unpack.
- ▌ PDF · nielsmadan bundleRead, extract, merge, split, rotate, watermark, create, fill forms, encrypt/decrypt, or OCR PDF files. Use whenever a `.pdf` is mentioned or a PDF is requested.
- ▌ Plan · nielsmadanLightweight planning workflow — delegate read-only planning to a Fable subagent, then implement in auto mode after a single go-ahead gate. The middle tier between "just do it" (tiny tasks) and heavyweight superpowers planning (big features). Never enters plan mode, so it sidesteps the plan-mode permission prompts. Use when the user invokes /plan, or wants a plan for a medium-sized task before implementing.
- ▌ Temp · nielsmadanMake temporary code changes for testing that can be easily undone. Use when user says "temp", "temporary change", "temporarily enable/disable/show/hide", or needs to force a UI state, bypass a guard, or flip a feature flag for local testing. Also handles "temp undo" to revert all temporary changes, and "temp list" to show them.
- ▌ Test · nielsmadan bundleAssess test state and run the right action (default, no args): run the suite, then find failures, coverage gaps, and quality issues and route into fix, generate, or review. Explicit modes: review (--review, check test quality), generate (--generate <target>, create tests). Scope: --staged, --changed, --all, or context-based. Use when unsure what the tests need, or for test quality and creation.
- ▌ Guide · nielsmadanWalk the user through a multi-step task (e.g. cloud console / permission / dashboard setup) with a live step tracker that is re-printed at the bottom of every reply so they never scroll up. Use when the user asks to "guide me through", "walk me through", "give me step by step" instructions, "how do I set up ..." for a UI/console task, or invokes guide. Also use when, mid-guide, they say a step "isn't working", "the menu isn't there", or ask a clarifying question about a step.
- ▌ Commit · nielsmadan bundleCommit ONLY the changes this session made, never another agent's work in the same checkout. With a message argument, makes one commit; with no argument, splits the session's work into the fewest self-contained commits and generates a short feat/fix/chore message for each. Use when the user says "commit", "commit this", "commit my changes", "commit just my/this session's changes", or invokes commit — especially when multiple agents share one working tree.
- ▌ Deslop · nielsmadan bundleCopy-edit text to strip AI/LLM writing tells ("slop") and make it read as human-written — overused words (delve, showcase, robust), significance-inflation phrases ("stands as a testament to", "plays a pivotal role"), scene-setting openers ("in today's fast-paced world"), hedging, em-dash overuse, rule-of-three, and "it's not X, it's Y" parallelism. Use when asked to "deslop", "de-slop", "remove AI tells", "make this sound less like AI/ChatGPT", "make this sound human", or copy-edit a draft (.md, .txt, prose, emails, docs) that reads as machine-generated.
- ▌ Status · nielsmadanReport the current coding-session status: summarize staged, unstaged, and untracked files; identify the last task and whether its work is present or committed; report code-review coverage; summarize active plans or review todos; list remaining session work; and recommend the next task. Use when the user invokes status or asks 'where are we?', 'what changed?', 'what is left?', or 'what should I do next?' about the current coding session. Do not use for service health or deployment status.
- ▌ Explain · nielsmadanGenerate project explanation docs in docs/explain/ covering architecture, flows, syntax, system APIs, infra, and testing, or explain a code diff, commit, branch, or PR directly in the conversation with --diff.
- ▌ Hard Fix · nielsmadan bundleEscalation workflow for stubborn bugs. Use when a bug persists after multiple fix attempts, you've tried several approaches, or you're stuck.
- ▌ Ideation · nielsmadan bundleGenerate ideas with structure when you're stumped — on what to build next, what the real problem is, or how to solve it. Pulls in research for context, diverges wide using matched frameworks, then converges on a prioritized few. Use when the user says "I'm stuck", "I'm stumped", "ideate", "brainstorm ideas", "help me think of", "what could I add", "what should I build", "I don't know what the problem is", "how could I solve", or wants idea generation on any topic (product, technical, business, writing, personal). For auditing an existing product against its users, use review-product instead.
- ▌ Longshot · nielsmadan bundleRun a long autonomous build session from a brief — interrogate every open decision up front with `blind-spots`, then execute for hours without check-ins, deciding ambiguities as recorded rulings instead of stopping to ask. Per task a fresh implementer subagent, an independent spec+quality reviewer, a fix loop, then a whole-branch review and a handoff report with rulings, deferred questions and a squash proposal. Use when the user says "longshot", "work on this independently", "run with this", "ask me everything up front then go", "I'll be away", or hands over a multi-hour feature or package to build end to end. Do NOT use for a single small change — use `plan` for that.
- ▌ Debug Log · nielsmadanAdd debug logging to trace code execution. Use when debugging, tracing control flow, or investigating unexpected behavior. Any language (JS/TS, Python, Go).
- ▌ Perf Test · nielsmadan bundleSet up and run performance tests (profiling, load testing, or E2E scenarios), preserving occasional manual procedures and results for future repetition. Use when measuring code or endpoint performance, or before/after optimization work.
- ▌ Read Docs · nielsmadanSearch internal project docs (docs/, README.md, CLAUDE.md, *.md) for patterns, conventions, architecture. Use proactively before features, in new code areas, or when debugging.
- ▌ Blind Spots · nielsmadan bundleSurface the decisions a plan or design left silently assumed, by interviewing the user in dependency order until nothing is unstated. Use when the user says "blind spots", "grill me", "stress-test this plan", "poke holes in this", "what am I missing", "interrogate this design", "what haven't I decided", or hands over a loose idea to sharpen before building. Do NOT use to explain a previous reply in more detail (use `huh`), to generate ideas when it is unclear what to build (use `ideation`), or to review an already-written plan with agents (use `review-plan`).
- ▌ Code Review · nielsmadan bundleClean up code comments, then review code with comprehensive and quick modes. Use when reviewing code changes, PRs, or specific files for quality, bugs, and best practices.
- ▌ Review Logs · nielsmadan bundleAnalyze Claude Code session transcripts for failure patterns and suggest fixes. Triggers "review logs", "session analysis", "failure patterns".
- ▌ Review Perf · nielsmadan bundlePerformance analysis for algorithmic complexity, memory leaks, N+1 queries, and render issues. Use when code feels slow, after adding loops/queries, or before scaling up.
- ▌ Review Plan · nielsmadan bundleMulti-agent review of implementation plans. Use after creating a plan but before implementing, especially for complex or risky changes.
- ▌ Review Todo · nielsmadanTurn a completed code review into a persistent workflow that immediately proposes the complete ordered implementation plan, waits for whole-plan approval, then implements and commits every accepted finding. Use after a review when the user invokes review-todo, says "turn this review into a todo list", "work through these review findings", supplies directives such as "fix 2, 3; ignore 7", asks to revise or approve the plan, or resumes review work. Do not use to perform the original code review.
- ▌ Library Docs · nielsmadan bundleGenerate and refresh a per-repo `library-use` reference — official docs, changelog links, pinned versions, and distilled correct-usage conventions for the repo's fast-moving / niche libraries. Use when the user says "library-docs", "doc links", "document the libraries we use", "generate/refresh library docs", "library conventions", or wants a per-repo doc-links reference the agent (and `review-library-use`) can consult. On re-run it version-checks every entry and updates it.
- ▌ Pre Existing · nielsmadanFix any test/lint/type/build/CI failure instead of dismissing it as pre-existing, flaky, or unrelated. Triggers on red checks or `/pre-existing`.
- ▌ Review Swift · nielsmadan bundleSwift-specific code review focused on JUDGMENT-level design a linter and the compiler can't decide — state modeling with enums and value types (make invalid states unrepresentable), optional and error modeling, concurrency isolation intent, ARC ownership, SwiftUI identity/lifetime/dependencies, and escape hatches (`!`, `as!`, `try!`, `@unchecked Sendable`) that compile but hide a modeling problem. Deliberately does NOT duplicate SwiftLint, swift-format, or Swift 6 strict-concurrency diagnostics. Auto-invoked by `code-review` on Swift projects. Triggers "review swift", "swift review", "swiftui review", "swift concurrency review".
- ▌ Evaluate Tech · nielsmadan bundleStructured evaluation before adopting a library, tool, or hosted service — enumerate candidates wide, then score every one against an identical rubric where maintenance health is a mandatory gate, not an afterthought. Use when the user asks "which library/package should I use", "what should we use for X", "which service/vendor should we pick", "is this package still maintained", "alternatives to X", "should we add this dependency", "should we switch to X", "pick a tool/service for", "evaluate this dependency/tool/service", or is choosing between named options to adopt. Covers code dependencies (npm, pip, pub, cargo, go), CLI/dev tools and dev software, and SaaS/API/hosted services. For learning how to USE something already chosen, or for open-ended technical research, use research-tech instead.
- ▌ Research Tech · nielsmadan bundleResearch any technical / developer topic online — libraries, errors, best-practice/how-to, tool·library·model comparisons, product capabilities, and ecosystem/community signal. Use when you'll act on the answer as a developer (write code, debug, choose a tool). For non-technical topics (science, history, consumer, personal, fact-check) use `research-general`.
- ▌ Skill Creator · nielsmadan bundleGuide for creating or improving Claude skills. Triggers "create a skill", "build a skill", "new skill", "improve this skill", "skill for [use case]".
- ▌ Review History · nielsmadanAnalyze how code changed over time. Use when investigating regressions, understanding why code was written a certain way, or finding when a behavior changed.
- ▌ Review Product · nielsmadan bundleReview a product from the user's perspective — build/refine a user persona, map their use cases (jobs-to-be-done), then audit the product for friction, gaps, and things to add or change. Triggers "review product", "product review", "review from the user's perspective", "product/UX critique", "what's missing for users". Use --live to also exercise the running app.
- ▌ Second Opinion · nielsmadanGet external AI opinions on a problem or question. Use when you want diverse perspectives from the agent CLIs you are not running (Claude, Codex, Pi, OpenCode).
- ▌ Squash Commits · nielsmadanSquash unpushed commits into clean, higher-level feat/fix/chore commits that follow the project commit policy. Use when local history has too many small/WIP/fixup commits (common after superhuman or gsd runs), or the user says "squash commits", "squash my commits", "tidy history", "clean up commits", "combine commits before pushing".
- ▌ Review Comments · nielsmadanReview and clean up code comments for necessity, accuracy, non-duplication, clarity, and concise explanation of rationale. Use when comments may restate code, repeat a "why" already explained in the file, contain stale or vague claims, or need tightening before a PR.
- ▌ Review Security · nielsmadan bundleSecurity audit for vulnerabilities, secrets, and unsafe patterns. Use before releases, after adding auth code, or when reviewing third-party integrations.
- ▌ Check Agent Logs · nielsmadan bundleSearch past Claude Code, Codex, OpenCode, and Pi session logs to recover context from earlier work across the current project and sibling checkouts. Use when a bug, topic, or decision was handled in a previous session but its agent or checkout is unknown; when the user says "we fixed/discussed this before", "find the session where", "recover prior context", "check agent logs", "check Claude projects", "search past sessions/transcripts", or "which checkout was that in". Do not use for aggregate failure-pattern analysis; use review-logs for that.
- ▌ Research General · nielsmadan bundleResearch a non-technical topic online (science, history, news, policy, regional/regulatory, consumer purchases, personal decisions, fact-check). The default for research in non-code repos (e.g. a notes vault). For technical/developer topics — libraries, errors, tooling, and even choosing/evaluating dev tools or products — use `research-tech`.
- ▌ Review Cleancode · nielsmadan bundleReview code for clean-code principles — SOLID, DRY, YAGNI, KISS, code smells. Triggers "review clean code", "check DRY/SOLID", "code smells".
- ▌ Resolve Conflicts · nielsmadanResolve git conflicts from any operation (merge, rebase, cherry-pick, stash, revert). Use when encountering conflicted files during git operations.
- ▌ Review Interfaces · nielsmadan bundleReview interface design for functions, classes, modules, components — naming, params, encapsulation, YAGNI, usability. Triggers "review interfaces".
- ▌ Review Typescript · nielsmadan bundleTypeScript-specific code review focused on JUDGMENT-level type design a linter can't decide — type modeling (make invalid states unrepresentable), inference-vs-annotation calls, and casts/`any` that hide a real modeling problem. Deliberately does NOT duplicate typescript-eslint. Auto-invoked by `code-review` on TypeScript projects. Triggers "review typescript", "typescript review", "type design review".
- ▌ Review Library Use · nielsmadanReview code for correct use of the repo's third-party libraries — checks the scoped files against the version-specific conventions recorded in the repo's `library-use` reference (docs-derived correct-usage rules, API contracts, footguns). Catches stale-API usage, deprecated patterns, and doc-violating misuse a general reviewer misses. Auto-invoked by `code-review` when a `library-use` reference exists. Triggers "review library use", "check library usage", "are we using this library correctly", "library convention review".
- ▌ Review Architecture · nielsmadanReview system architecture — layering, module boundaries, coupling/cohesion, pattern fit, quality attributes (scalability, resilience, evolvability), and architectural smells. Triggers "review architecture", "architecture review", "system design review", "check architecture".
- ▌ Improve Agent Instructions · nielsmadan bundleAudit and improve the always-loaded agent instruction files (AGENTS.md, CLAUDE.md, GEMINI.md). Triggers "check/audit/update/improve/fix/revise CLAUDE.md or AGENTS.md", "project memory optimization", "trim my instruction file". Not for general docs.